def _createCoreHumans(image: ImageForRedetection) -> List[Human]:
"""
Create core humans for redetection
Args:
image: image and bounding boxes for redetection
Returns:
Human object list. one object for one bbox
"""
humans = [Human() for _ in range(len(image.bBoxes))]
for index, human in enumerate(humans):
human.img = image.image.coreImage
human.detection.rect = image.bBoxes[index].coreRectF
human.detection.score = 1
return humans
def createHumanDetection(image: VLImage, detection: Detection,
landmarks17: Optional[CoreLandmarks17]):
"""
Create human detection structure from core detection result.
Args:
image: origin image
detection: sdk human detection
landmarks17: estimated landsmarks17
Returns:
HumanDetection structure
"""
human = Human()
human.img = image.coreImage
human.detection = detection
if landmarks17:
human.landmarks17_opt.set(landmarks17)
return HumanDetection(human, image)
def detectOne(
self, image: VLImage, detectArea: Optional[Rect] = None, limit: int = 5, detectLandmarks: bool = True
) -> Union[None, HumanDetection]:
"""
Detect just one best detection on the image.
Args:
image: image. Format must be R8G8B8
detectArea: rectangle area which contains human to detect. If not set will be set image.rect
limit: max number of detections for input image
detectLandmarks: detect or not landmarks
Returns:
human detection if human is found otherwise None
Raises:
LunaSDKException: if detectOne is failed or image format has wrong the format
"""
assertImageForDetection(image)
detectionType = self._getDetectionType(detectLandmarks)
if detectArea is None:
forDetection = ImageForDetection(image=image, detectArea=image.rect)
else:
forDetection = ImageForDetection(image=image, detectArea=detectArea)
imgs, detectAreas = getArgsForCoreDetectorForImages([forDetection])
error, detectRes = self._detector.detect(imgs, detectAreas, limit, detectionType)
assertError(error)
detections = detectRes.getDetections(0)
landmarks17Array = detectRes.getLandmarks17(0)
isReplyNotAssumesDetection = detectRes.getSize() == 1
if isReplyNotAssumesDetection:
isDetectionExistsNValid = len(detections) != 0 and detections[0].isValid()
if not isDetectionExistsNValid:
return None
human = Human()
human.img = image.coreImage
human.detection = detections[0]
if landmarks17Array and landmarks17Array[0] is not None:
human.landmarks17_opt.set(landmarks17Array[0])
return HumanDetection(human, image)
def detect(
self, images: List[Union[VLImage, ImageForDetection]], limit: int = 5, detectLandmarks: bool = True
) -> List[List[HumanDetection]]:
"""
Batch detect human bodies on images.
Args:
images: input images list. Format must be R8G8B8
limit: max number of detections per input image
detectLandmarks: detect or not landmarks
Returns:
return list of lists detection, order of detection lists is corresponding to order input images
"""
coreImages, detectAreas = getArgsForCoreDetectorForImages(images)
detectionType = self._getDetectionType(detectLandmarks)
validateBatchDetectInput(self._detector, coreImages, detectAreas)
error, fsdkDetectRes = self._detector.detect(coreImages, detectAreas, limit, detectionType)
assertError(error)
res = []
for imageIdx in range(fsdkDetectRes.getSize()):
imagesDetections = []
detections = fsdkDetectRes.getDetections(imageIdx)
landmarks17Array = fsdkDetectRes.getLandmarks17(imageIdx)
for idx, detection in enumerate(detections):
human = Human()
human.img = coreImages[imageIdx]
human.detection = detection
if detectLandmarks:
human.landmarks17_opt.set(landmarks17Array[idx])
imagesDetections.append(human)
image = images[imageIdx]
vlImage = image if isinstance(image, VLImage) else image.image
res.append([HumanDetection(human, vlImage) for human in imagesDetections])
return res